Long-term Risk of Upper Gastrointestinal Bleeding After Helicobacter pylori Eradication

A Population-based Cohort Study

Fang Jiang; Chuan-Guo Guo; Ka Shing Cheung; Wai K. Leung


Aliment Pharmacol Ther. 2021;54(9):1162-1169. 

In This Article


Data Source

This was a retrospective cohort study. All data were retrieved from the Clinical Data Analysis and Reporting System (CDARS), which is managed by Hong Kong Hospital Authority. The Hong Kong Hospital Authority is a publicly funded organisation to provide medical services to 7.3 million people, including 90% of primary, secondary and tertiary care services. Patient's clinical information including diagnoses, prescription and dispensing records, hospitalisation and laboratory results from all public hospitals and clinics were recorded in the CDARS and the details have been described in our previous studies.[4,14–19] Data in the CDARS are anonymised to protect patient's privacy, and a unique reference key is used to match every single patient record for subsequent analysis. The International Classification of Diseases, 9th Revision (ICD-9), code was used for our study. The accuracy of the codes for GIB in the CDARS has been verified in our previous study.[17] This study was approved by the Institutional Review Board of the University of Hong Kong and the Hospital Authority Hong Kong West Cluster (Reference Number: UW 16–545 and UW 20–152).

Study Subjects

We compared the risks of UGIB of two large cohorts including a H. pylori eradicated cohort and a H. pylori-negative cohort. For the first cohort, we have previously identified all H. pylori-infected patients who had received the first course of clarithromycin-containing triple therapy for H. pylori between 2003 and 2012 in the CDARS.[14–16,18,19] We excluded patients who were younger than 18 years, with any gastrointestinal cancers, inflammatory bowel disease, bleeding tendency, gastroenteritis or colitis due to radiations, had prior excision of any gastrointestinal tract segment, oesophageal varices, retreatment for H. pylori after the index eradication therapy, and death before the start date. Although the outcome of eradication therapy was usually monitored by urea breath test or endoscopy-based test in Hong Kong, post-eradication H. pylori status was not available in the CDARS. To identify patients who had failed initial H. pylori eradication, we used the need of retreatment after the initial eradication therapy as the marker of eradication failure. These retreatment regimes for H. pylori included (a) repeated course of clarithromycin-based triple therapy; (b) second-line therapy with bismuth-based quadruple therapy or levofloxacin-containing triple therapy; or (c) rifabutin-based therapy.

As a control group, we included another cohort of H. pylori-negative patients who underwent the first upper endoscopy during the same period and were found to have no H. pylori on histological examination. Moreover, these patients should not have received any H. pylori eradication therapy, both before index endoscopy and during the long-term follow-up. The details of the patient selection in the two cohorts are depicted in Figure S1.

Exposure of Interest and Covariates

Potential risk factors for UGIB included patient's sex, age of receiving triple therapy or upper endoscopy, past history of UGIB, peptic ulcer, other medical illnesses (diabetes mellitus, hypertension, intracranial haemorrhage, stroke, ischaemic heart disease [IHD], chronic renal failure [CRF] and cirrhosis) as well as concurrent medication use (proton pump inhibitors [PPIs], histamine type 2 receptor antagonists [H2RA], non-steroidal anti-inflammatory drugs [NSAIDs], aspirin, anticoagulants, other antiplatelet agents, selective serotonin reuptake inhibitors [SSRI] and bisphosphonate). New prescriptions for PPI and H2RA drugs in the last 6 weeks of the event date or censor date were not counted to reduce potential indication bias. The ICD-9 codes for all these medical conditions are shown in Table S1.

Propensity Score Matching

Propensity score (PS) matching analysis in a 1:1 ratio was conducted to balance the differences in baseline characteristics between the H. pylori eradicated group and the H. pylori-negative group. The PS matching included patient's sex, age, year of H. pylori eradication therapy or upper endoscopy, baseline medical illnesses and concurrent medication use in the first 3-monthly intervals after the start date. An absolute standardised difference (ASD) <0.10 indicates a good balance of these covariates between the two groups.

Study Outcome

The primary endpoint was hospitalisation for non-variceal UGIB. All patients were observed from the start date, which was 60 days after H. pylori therapy (H. pylori eradicated cohort) or upper endoscopy (H. pylori-negative cohort) until the occurrence of UGIB, death or the end of study (30 June 2016). UGIB that occurred within 60 days from the start date of eradication therapy or endoscopy were not counted as this may be related to unresolved event leading to H. pylori eradication and endoscopy. The ICD-9 codes for identifying UGIB are shown in Table S1.

Statistical Analysis

Continuous variables were presented as median with interquartile range (IQR), whereas categorical variables were expressed as frequencies and percentages (%). Multivariable Cox proportional hazards regression model after PS matching was conducted to compare the risks of UGIB as the primary model and expressed as hazard ratio (HR) with 95% confidence intervals (CIs). The PS matching adjusting for time-varying concomitant medications and univariable Cox regression model after PS matching were also conducted as sensitivity analyses. Concurrent medication uses during follow-up were treated as time-varying covariates and the follow-up period was split into 3-monthly intervals with drug usage defined as at least 7-day use in each interval.

Kaplan-Meier curve with log-rank test was performed to compare the proportion of patients with UGIB in the H. pylori eradicated group and H. pylori-negative group. Subgroup analyses were performed to determine the risks of UGIB in the first two-year and in the subsequent years. Another subgroup analysis was also performed to compare risks of UGIB after stratified by different age groups (18-45, 45–60, 60–75, 75 years and older).

To account for the effects of prior UGIB or peptic ulcer on the risk of subsequent UGIB, sensitivity analyses were performed after excluding and including only patients with a history of prior UGIB alone, and prior UGIB or peptic ulcer. A sensitivity analysis using peptic ulcer bleeding as an outcome was conducted, with further subgroup analyses in the first two-year and in the subsequent years. All statistical tests were two-sided and a P-value of < .05 was considered statistically significant. Statistical analyses were conducted by the R statistical software version 4.0.3 (R Foundation for Statistical Computing, Vienna, Austria, 2019).